The X-ray source in non-magnetic cataclysmic variables

Transcription

1 A&A manuscript no. (will be inserted by hand later) Your thesaurus codes are: 06( ; ; ) ASTRONOMY AND ASTROPHYSICS The X-ray source in non-magnetic cataclysmic variables André van Teeseling 1, Klaus Beuermann 1, and Frank Verbunt 2 1 Universitäts-Sternwarte Göttingen, Geismarlandstr. 11, Göttingen, Germany 2 Astronomical Institute, Utrecht University, P.O.Box , 3508 TA Utrecht, The Netherlands Received 13 December 1995 / Accepted 13 April 1996 Abstract. We have analysed all available rosat pspc observations of non-magnetic cataclysmic variables. The VY Scl and non-su UMa systems have harder X-ray spectra than the SU UMa and UX UMa systems. The old nova RR Pic has a very soft X-ray spectrum and might not be an ordinary nonmagnetic cataclysmic variable. There appears to be no clear correlation of the X-ray temperature, the emission measure, or the X-ray luminosity with the accretion rate. However, we have found an anti-correlation between the observable emission measure and the orbital inclination of the system, from which we conclude that the X-ray emitting region is close to the white dwarf and has a size smaller than the white dwarf radius. Key words: accretion, accretion disks - stars: novae, cataclysmic variables - X-rays: stars 1. Introduction In cataclysmic variables a white dwarf accretes matter from a low-mass companion star. If the white dwarf has a weak magnetic field (B < 10 6 G) the accreted matter forms an accretion disk which may continue to the surface of the white dwarf. The most popular theory to explain the X-rays from non-magnetic cataclysmic variables is that the X-rays are emitted by the boundary layer between the white dwarf and the disk. In this boundary layer the matter decelerates from Keplerian velocities to the rotation velocity of the white dwarf. In principle, as much as half of the total accretion energy could be emitted by the boundary layer. Pringle & Savonije (1979) argued that when the mass accretion rate is low ( : m g s?1 ) the boundary is optically thin and is heated to temperatures of 10 8 K. If the accretion rate is high ( : m g s?1 ) the boundary layer would be optically thick and radiate 10 5 K blackbody-like radiation. There are several observational problems with this simple boundary-layer theory for the X-ray source in non-magnetic cataclysmic variables: Send offprint requests to: A. van Teeseling 1) A soft X-ray component could only be detected so far in 5 dwarf novae in outburst (Rappaport et al. 1974; Mason et al. 1978; Van der Woerd 1987; Wheatley et al. 1996). In three of them, VW Hyi, OY Car, and SS Cyg, the soft X-ray source is probably not completely optically thick (Van der Woerd 1987; Van Teeseling et al. 1993; Naylor et al. 1988; Mauche 1995; Mauche et al. 1995). In the other two, U Gem and Z Cam, the nature of the soft X-rays is not yet clear. No optically thick soft X-ray component could be detected in any nova-like variable (e.g. Van Teeseling et al. 1995). 2) Also in high-accretion-rate systems there is a hot optically thin X-ray source. This could be explained if the outer parts of the boundary layer remain optically thin (Patterson & Raymond 1985). 3) For high-accretion-rate systems the X-ray luminosity is much less than the ultraviolet+optical luminosity (Van Teeseling & Verbunt 1994). The bolometric X-ray luminosity does not seem to increase with increasing accretion rate. 4) There is evidence for orbital modulation in the X-rays from some non-eclipsing systems (Van Teeseling & Verbunt 1994; Van Teeseling et al. 1995). There is, however, also evidence in favour of X-rays related to the inner part of the accretion disk. The outburst behaviour of dwarf novae (Wheatley et al. 1995; Van Teeseling & Verbunt 1994; Ponman et al. 1995) shows that the X-rays rise back to the quiescence level when the optical outburst is completely over. Also, the X-rays in the SU UMa dwarf nova HT Cas eclipse together with the white dwarf (Wood et al. 1995a; Mukai et al. 1995), consistent with the X-rays originating from the immediate neighborhood of the white dwarf. In this article we analyse rosat pspc archive data of nonmagnetic cataclysmic variables. With a large sample we investigate how these observations constrain possible theories about the origin of the X-rays from non-magnetic cataclysmic variables. 2. Observations We have extracted all available pspc observations of nonmagnetic cataclysmic variables out of the rosat public archive at the Max-Planck Institute for Extraterrestrial Physics in

2 2 A. van Teeseling et al.: The X-ray source in non-magnetic cataclysmic variables Garching. All observations were analysed using the EXSAS software package (Zimmermann et al. 1994) as described in Van Teeseling & Verbunt (1994). Briefly, first all contaminating sources with an existence probability higher than 99.75% were excised from the image. Then, cataclysmic-variable counts were extracted from a circle with a radius of , centered on the cataclysmic-variable position. The background count rate was determined from a ring between and , with the same center, and corrected for vignetting. For the off-axis observation of HT Cas we have used radii of , , and , respectively. To obtain 2 fits to the X-ray spectra, the counts in pulseheight channels 11 to 235 were binned in such a way that each bin had the same signal-to-noise ratio. If possible, the minimum accepted signal-to-noise ratio in each bin was 5. Because there are 5 to 6 independent energy bands in the pspc, we have taken a minimum of 6 bins and a maximum of 19 bins. We have also binned the observed spectra in photon-channel bandpasses A (channels 11 to 41; kev), C (52 to 90; kev), D (91 to 201; kev), and B = C + D. The rosat pspc bandpass covers only a limited range of energies, kev. Determination of bolometric X-ray temperatures and fluxes is therefore uncertain when the bulk of the flux is emitted at energies below or above this energy range. We will ignore this uncertainty, because we do not think that our conclusions are affected by it. In Table 1 we give a list of the observations together with the derived count rates and hardness ratios. The sample contains 6 UX UMa systems, 3 Z Cam systems, 3 U Gem systems, 5 VY Scl systems (or anti-dwarf novae), 2 non-su UMa systems, 16 SU UMa systems, 2 unclassified dwarf novae, 2 double-degenerate systems, and 3 additional old novae. We have not included outburst data of the dwarf novae WX Hyi (Van Teeseling & Verbunt 1994), SU UMa, and RU Peg (Silber et al. 1994), but selected the data obtained when these systems were in quiescence. It has been proposed that V426 Oph, SW UMa, V795 Her, TT Ari, and WZ Sge contain an intermediate polar (Szkody 1986; Shafter et al. 1986; Shafter et al. 1990; Patterson 1994). However, we will show that their X-ray properties, with perhaps the exception of V426 Oph, fit well into the X-ray properties of non-magnetic systems. Two of the three additional old novae were not detected. The third, RR Pic, might not be a non-magnetic cataclysmic variable. 3. X-ray spectra In Fig. 1 we have plotted the average hardness ratio (B? A)=(B + A), where A and B are shorthand notations for the count rates in bandpasses A and B, for all sources with a 1 error in the hardness ratio smaller than 0:2. Noteworthy is that all Z Cam, VY Scl, and non-su UMa systems have a hardness ratio higher than 0.5. Most SU UMa and UX UMa systems have much softer spectra. The old nova RR Pic has the softest spectrum. To get an impression of the X-ray spectra we have plotted in Fig. 2 the count ratios C=D vs. A=D in a diagram together Fig. 1. (B? A)=(B + A) hardness ratios of sources with a 1 error < 0:2. Plotted are different subclasses of cataclysmic variables with individual objects indicated as follows: =RR Pic; + =EI UMa; =GP Com and AM CVn with the ratios of various theoretical spectra (cf. Van Teeseling & Verbunt 1994). Absorption causes a spectrum to move to the left almost horizontally. All observed spectra lie in a narrow horizontal band consistent with absorbed optically thin spectra. All VY Scl and non-su UMa systems are on the left side of the band, a fact that is also illustrated in Fig. 1. This suggests that the X-ray spectra of these systems are highly absorbed or that they are not thermal bremsstrahlung spectra. This is confirmed by bremsstrahlung fits to their X-ray spectra: they have all hydrogen absorption columns n H > cm?2, in contrast to most of the SU UMa and UX UMa systems which have n H < cm?2. The inferred X-ray absorption columns depend on the assumed type of spectrum. Therefore, in Fig. 3 we compare the inferred n H columns with the reddening estimates from the 2200 Å feature in the iue spectra (Verbunt 1987). It appears that for all SU UMa, UX UMa, and Z Cam systems the inferred X-ray absorption column is consistent with the ultraviolet reddening. For all VY Scl and non-su UMa systems, however, the inferred X-ray absorption column is higher than the column derived from the ultraviolet reddening. This indicates that the high absorption in the X-ray spectra of these systems may be due to intrinsic absorption of the source or that their X-ray spectra are not simple thermal bremsstrahlung spectra. The old nova RR Pic falls completely outside the band formed by the other non-magnetic cataclysmic variables. This is illustrated in Fig. 4. An acceptable fit to the spectrum of RR Pic requires an additional soft component, e.g. a blackbody. With a hardness ratio of?0:49 0:08 RR Pic falls close to the region populated by AM Her systems (Beuermann & Thomas 1993) and may, in fact, be some type of magnetic cataclysmic variable. We note that RR Pic is one of the systems for which Verbunt (1987) found variability in the ultraviolet fluxes at a timescale shorter than the orbital period. In view

3 A. van Teeseling et al.: The X-ray source in non-magnetic cataclysmic variables 3 Table 1. Pointed rosat pspc observations of non-magnetic cataclysmic variables source type a P orb b date exp. (s) counts/s c (B? A)=(B + A) d references e V603 Aql sh Apr 6/ :252 0:005 0:83 0:03 TT Ari vy Aug 1/ :412 0:004 0:83 0:01 6,8 KR Aur vy Sep/Oct :071 0:002 0:96 0:05 13 TT Boo su Jul 16/ :013 0:002 0:37 0:18 BZ Cam vy Sep :081 0:004 0:95 0: Sep :068 0:004 0:83 0:08 1 HT Cas su Aug 17/ :098 0:004 0:48 0:04 10 WW Cet zc / :805 0:011 0:50 0:02 5 Z Cha su Apr 8/ :083 0:005 0:53 0:07 AC Cnc ux May :005 0:002?0:26 0:39 YZ Cnc su Apr :641 0:016 0:31 0: Oct 7/ :464 0:016 0:21 0:04 1 GP Com ac Dec :826 0:012 0:09 0:01 1 V394 Cra nr Mar < 0:005 AM CVn ac Dec 8/ :035 0:003 0:39 0:08 12 AB Dra zc Apr 6/ :204 0:005 0:84 0:03 5 AH Eri ug 1992 Mar 8/ :009 0:003 0:59 0:41 V795 Her sh Aug 11/ :034 0:002 0:81 0:07 11 VW Hyi su Jul 22/ :275 0:012 0:05 0:01 1,4 WX Hyi su Nov/Dec :515 0:008 0:46 0:02 1 T Leo su Jun :634 0:012 0:22 0:02 MV Lyr vy Nov 4/ :082 0:002 0:83 0:04 13 V2051 Oph ug? Sep :019 0:007 0:15 0: Sep :050 0:009 0:67 0:23 V426 Oph zc Mar 17/ :373 0:005 0:97 0:02 7 RR Pic nb Dec/1993 Apr :080 0:006?0:49 0:08 TY PsA su May 7/ :310 0:006 0:34 0:02 1 VZ Scl vy Dec :005 0:002 1:00 0:57 WZ Sge su Apr 10/ :291 0:004 0:21 0:02 9 V3885 Sgr ux Oct :163 0:004 0:38 0:03 1 RW Tri ux Aug 18/ :007 0:003 1:39 0:66 BZ UMa dn Oct :611 0:013 0:07 0: Oct 6/ :729 0:017 0:12 0:02 CI UMa dn May 11/ :004 0:002?0:19 0:44 CY UMa su Dec :096 0:004 0:03 0:04 1 DW UMa ux Oct :018 0:004 0:49 0:23 DV UMa su? Nov 9/ :005 0:003 0:20 0:53 EI Uma ug Oct 5/ :048 0:016?0:07 0:02 SU UMa su Apr 14/ :238 0:019 0:24 0:02 5 SW UMa su Apr/May :337 0:008 0:00 0: Oct 16/ :279 0:010?0:02 0:04 UX UMa ux Jan 3/ :018 0:003 0:06 0:19 2 SS UMi su Aug :049 0:003 0:69 0:08 CU Vel su May/Jun :151 0:005 0:25 0:03 1 IX Vel ux Apr 13/ :524 0:010 0:07 0:02 1, Apr :508 0:014 0:24 0:03 1, Oct 16/ :470 0:010 0:09 0:02 3 HV Vir su Jun 25/ :004 0:002 0:68 0:82 PW Vul na /1992 Mar 4767 < 0:005 a Subclass according to Ritter & Kolb (1993): ux = UX UMa nova-like variable, vy = VY Scl nova-like variable, sh = non-su UMa variable, su = SU UMa dwarf nova, zc = Z Cam dwarf nova, ug = U Gem dwarf nova, ac = AM CVn double degenerate, nr = recurrent nova, na = fast nova, nb = slow nova b Orbital period (days) c Time-averaged count rate in photon channels after background subtraction and vignetting correction; the error is dominated by the Poisson noise ; for dwarf novae only the quiescence count rate has been given d Hardness ratio with photon channel bands A = and B = after background subtraction e References: (1) Van Teeseling & Verbunt 1994; (2) Wood et al. 1995b; (3) Van Teeseling et al. 1995; (4) Belloni et al. 1991; (5) Vrtilek et al. 1994; (6) Baykal et al. 1995; (7) Rosen et al. 1994; (8) Robinson & Córdova 1992; (9) Richman & Patterson 1994; (10) Wood et al. 1995a; (11) Rosen et al. 1995; (12) Ulla 1995; (13) Schlegel & Singh 1995

4 4 A. van Teeseling et al.: The X-ray source in non-magnetic cataclysmic variables Fig. 2. rosat pspc count ratios of non-magnetic cataclysmic variables, with 1- error bars. Systems are identified with the first letters of their name. Different symbols are used for subclasses:? = UX UMa, = SU UMa, = Z Cam, and = other. Also plotted are the predicted count ratios for Mewe et al. spectra (Kaastra & Mewe 1993), blackbody spectra and thermal bremsstrahlung spectra. The theoretical spectra form bands with the solid boundary corresponding to zero interstellar absorption and the dashed boundary corresponding to a hydrogen column density of n H = cm?2. Absorption moves the predicted spectra almost horizontally to the left. The difference (in kev or log T ) between two subsequent marked points is constant. The highest blackbody temperature is 10 kev and the highest bremsstrahlung temperature is 90 kev. There is a 5% uncertainty in the theoretical curves due to uncertainties in the detector response matrix of the softness of the X-ray spectrum of RR Pic we have redetermined the reddening towards this system from the 2200 Å feature, using iue spectra SWP6625 and LWR5687. We find E B?V < 0:03 which is consistent with the n H cm?2 that we find from a bremsstrahlung+blackbody fit; the finite value of E B?V = 0:05 0:03 listed by Verbunt (1987) was based on spectra of lesser quality. 4. Emission measure and orbital inclination The emission measure is defined as EM = Z n 2 edv (1) Fig. 3. Inferred bremsstrahlung n H columns compared with E B?V derived from the 2200 Å feature in the iue spectra. Different symbols are used for subclasses:? = UX UMa, = SU UMa, = Z Cam, and = VY Scl or non-su UMa. The dotted line is n H = 5: E B?Vcm?2 (Predehl & Schmitt 1995) where n e is the electron density and V the emitting volume. In Fig. 5 we have plotted the observable emission measure against the orbital inclination, for all systems for which a reliable inclination is known and for which the distance is given by Warner (1987). There is a 99.9% significant anticorrelation: the emission measure is smaller for high-inclination

5 A. van Teeseling et al.: The X-ray source in non-magnetic cataclysmic variables 5 Fig. 4. Same as Fig. 2, but with a larger range plotted systems. This result does not strongly depend on the assumption that the X-ray spectrum is a thermal bremsstrahlung spectrum: the same anti-correlation is found if we use the quantity countrate distance 2, which is independent of the assumed spectrum. The anti-correlation is also present when distances as determined by Patterson (1984) are used. We have to beware of possible selection effects that may cause the anti-correlation. Of many sources in our sample the inclination is unknown, and even those which are known are sometimes uncertain. Especially, small inclinations are difficult to determine. In Fig. 5 there are no dwarf novae with an inclination < 30. It could be possible that these low-inclination dwarf novae have relatively low emission measures and fill in the lower left corner of Fig. 5. However, if we only consider the SU UMa dwarf novae in Fig. 5 the anti-correlation is still significant at a 99% level. The reality of the anti-correlation is also confirmed by the fact that the distance-corrected count rate is smallest for all eclipsing systems. There are no highinclination systems with high count rates or emission measures. For the high-inclination systems UX UMa, RW Tri, VZ Scl, DW UMa, and V2051 Oph we could not determine a reliable emission measure, because the exposures are low and the inferred spectral parameters are uncertain. We cannot exclude that these systems have high emission measures. However, the low count rates of these systems suggest that they fit nicely into the anti-correlation in Fig. 5. If we only include systems with accurate emission measures and inclinations, the anti-correlation is highly significant. The anti-correlation between emission measure and inclination according to Fig. 5 includes systems of different subclasses of cataclysmic variables: non-magnetic systems and candidate intermediate polars, and of very different accretion rates. This implies that the X-ray emission measure is not a strong function of the accretion rate, which is confirmed in the next Section. There is already a substantial decrease of the emission measure for moderate inclinations. Without the eclipsing systems with i > 70 the anti-correlation is still 99% significant. For Fig. 5. Count rate ((distance/100 pc) 2 ) and observable emission measure as a function of the orbital inclination. The errors in the corrected count rate and emission measure do not include the uncertainty in the distance. Different symbols are used for subclasses:? = UX UMa, = SU UMa, = Z Cam, and = other. The dashed line in the lower panel is equation 2 inclinations i < 70 the dependence of emission measure on inclination cannot be due to obscuration of the X-ray source by matter in the outer parts of the accretion disk or by the secondary. The anti-correlation between emission measure and inclination excludes all models in which the X-rays are emitted in a relatively large optically thin volume. On the other hand, if the X-rays originate from the inner part of the accretion disk and the scale height of the optically thin X-ray source is not much higher than the disk thickness, the observable emission measure could depend on the inclination. For example, an optically thin boundary layer with radial thickness d, which is partly obscured by the optically thick inner part of the disk with vertical thickness h, will have EM / 1= tan i, for tan i > d=h. If we assume such relation to be valid Fig. 5 indicates that numerically EM = tan i (cm?3 ) (2)

6 6 A. van Teeseling et al.: The X-ray source in non-magnetic cataclysmic variables A substantial decrease of the emission measure for moderate inclinations indicates that d < h R wd, with R wd the whitedwarf radius. This model would imply: In high-inclination systems most of the X-ray flux is absorbed by the accretion disk. We note that d and h probably also depend on the accretion rate. 5. X-ray and ultraviolet-optical flux With a limited sample of ten cataclysmic variables Van Teeseling & Verbunt (1994) showed that the ratio of X-ray to ultraviolet flux decreases with increasing orbital period. Their sample did not include any U Gem or Z Cam system. To investigate whether this relation holds for all non-magnetic cataclysmic variables we have plotted in Fig. 6 the ratio of bolometric X-ray to ultraviolet+optical flux of all systems with a reliable estimate of the X-ray flux. The ultraviolet+optical flux has been calculated with F uv+opt = 718f f f f f 5500 (3) The ultraviolet fluxes have been taken from Verbunt (1987), supplemented with some more recent iue observations. If no iue observations are available the uv+optical flux has been estimated from the optical magnitude using log F uv+opt =?0:4m v? 4:32. Both the X-ray and the ultraviolet flux have been corrected for absorption in the line of sight. Intrinsic errors in the derived ratios are caused by three uncertainties: First, some of the rosat observations cover only a small part of the orbital period. If the exposure time is short and the X-ray count rate is strongly variable, this could result in an observed count rate which significantly deviates from the real average count rate (e.g. Van Teeseling & Verbunt 1994). Second, the ultraviolet observations and some of the optical magnitudes are not simultaneous with the rosat observations. If the uv or optical flux varies strongly with time, this could give a wrong ratio of X-ray to uv+optical flux. Third, we have derived bolometric X-ray fluxes from measurements in the limited rosat bandpass of kev. However, we estimate that these uncertainties are much smaller than the range of the X-ray-to-ultraviolet flux ratio, which spans more than three orders of magnitude. We therefore do not think that these uncertainties affect our conclusions. From Fig. 6 it is clear that the relation found by Van Teeseling & Verbunt (1994) is more generally true. All SU UMa systems have a ratio near 0.1, all VY Scl and non-su UMa systems have a ratio near 0.01, and all UX UMa systems have a ratio < 0:001. However, the Z Cam systems V426 Oph and WW Cet have ratios of 0:05 and 0:5, respectively, and the U Gem system EI UMa has a ratio of 1. As argued by Patterson & Raymond (1985) and Van Teeseling & Verbunt (1994) the anti-correlation between the ratio of X-ray to uv+optical flux and the orbital period is probably due to an increasing mass accretion rate with increasing orbital period. The deviating ratios of V426 Oph, WW Cet, and EI UMa could be explained Fig. 6. The top panel shows the ratio of X-ray flux to ultraviolet+optical flux. For the open symbols the ratio has been derived from the optical magnitude only. The middle and bottom panels show the bolometric X-ray luminosity and the ( Å) uv luminosity. The uv flux is given by Verbunt (1987) and the distances by Warner (1987). Different symbols are used for subclasses:? = UX UMa, = SU UMa, = U Gem, = Z Cam, and = other if these systems have (temporarily) lower mass accretion rates than the UX UMa systems. This is confirmed by the ultraviolet luminosities of V426 Oph and WW Cet, which are relatively low compared to the ultraviolet luminosities of the UX UMa systems. Another possibility is that these systems are intermediate polars or DQ Her systems. Most DQ Her systems appear to have a higher ratio of X-ray to uv+optical flux than nonmagnetic systems with the same period and fill up the upper right corner of the plot. Patterson (1994) lists V426 Oph as a poor candidate member of the DQ Her systems.

7 A. van Teeseling et al.: The X-ray source in non-magnetic cataclysmic variables 7 There is another exception not shown in Fig. 6: the doubledegenerate system AM CVn. The low X-ray luminosity of AM CVn has been noticed by Ulla (1995). With an orbital period of 0:29 hours it has an unexpectedly small ratio of 0:002. Although there is no indisputable evidence that AM CVn is a double-degenerate binary, such a model does explain its properties rather well (Patterson et al. 1992; Provencal et al. 1995). The small ratio of X-ray to uv+optical flux may be explained as due to a high accretion rate in AM CVn, comparable to the UX UMa s. This is consistent with the presence of an optically thick accretion disk (Patterson et al. 1992), as well as the theoretical prediction that for a degenerate donor star the shortest orbital period corresponds to the highest mass transfer rate (Vila 1971). In Fig. 6 we have also plotted the bolometric X-ray luminosity and the Å ultraviolet luminosity for systems with a distance given by Warner (1987). The X-ray luminosity shows no correlation with the orbital period, in agreement with the anti-correlation between the emission measure and the orbital inclination. The decreasing ratio of X-ray to uv+optical flux is entirely due to an increasing uv luminosity with increasing period. The spread in the X-ray and uv luminosity is a factor of ten larger then the spread in the ratio of X-ray to uv+optical flux. This is illustrated by UX UMa and IX Vel. UX UMa has an X-ray luminosity six times smaller then IX Vel, but also its uv luminosity is a factor of six smaller. The additional spread in the X-ray luminosity is at least partly due to different orbital inclinations (see Fig. 5). However, in the luminosity there is an additional error due to the uncertainty in the distance, which may be a factor of two. 6. X-ray temperature and accretion rate According to the standard boundary-layer theory (Pringle & Savonije 1979; Tylenda 1981) low-accretion-rate systems have a higher X-ray temperature than high-accretion-rate systems. In Fig. 7 we have plotted the bremsstrahlung temperatures as a function of the ultraviolet+optical luminosity for systems with a distance given by Warner (1987). The accretion rate should be roughly proportional to the ultraviolet+optical luminosity. The luminosity has been determined as described in Sect. 5. The temperature has been determined from a thermal bremsstrahlung fit without adding an extra emission line near 1 kev, even when the fit was formally unacceptable. The high-luminosity UX UMa systems have a lower average temperature than the systems with lower luminosities. However, there is no significant anti-correlation between the bremsstrahlung temperature and the uv+optical luminosity. A relation T brem / M :?1 (Tylenda 1981) is certainly not consistent with the observations. This is reminiscent of extremeultraviolet/soft-x-ray observations of SS Cyg and VW Hyi in outburst in which the hardness of the spectra remained practically constant while the count rate increased by 2 orders of magnitude (Mauche et al. 1995; Van der Woerd & Heise 1987). Van Teeseling & Verbunt (1994), however, found that the temperature of the X-ray source in WX Hyi significantly Fig. 7. Bremsstrahlung temperature as a function of the ultraviolet+optical luminosity. Different symbols are used for subclasses:? = UX UMa, = SU UMa, and = other. For the open symbols the luminosity has been derived from the optical magnitude only decreased during outburst, although the hardness ratio did not vary significantly. Patterson & Raymond (1985) also found that the temperatures of high-accretion-rate systems were higher than the most simple boundary-layer theories predict. They attributed this to a density gradient in the optically thick boundary layer, such that there is always a hot optically thin layer which emits hard X- rays. With more recent observations available (e.g. Van Teeseling et al. 1995) this seems unlikely, because there is no observational evidence for a hot optically thick boundary layer in these systems. 7. Conclusions There appears to be no correlation of the X-ray temperature, the emission measure, or the X-ray luminosity with the accretion rate. This is difficult to explain when the X-rays are emitted by the boundary layer. We confirm that the ratio of X-ray to ultraviolet+optical flux decreases with increasing orbital period, with the exception of AM CVn and three long-period systems. This implies that the ratio of X-ray to ultraviolet+optical flux is anti-correlated with the accretion rate. Again, this is not consistent with the most simple boundary-layer models. There is strong evidence for an anti-correlation between the observable emission measure and the orbital inclination. This suggests that most of the X-rays are emitted very close to the white dwarf. We can exclude that most X-rays are emitted by an optically thin cloud with a scale height much larger than the white-dwarf radius. Acknowledgements. This research was supported by the Deutsche Agentur für Raumfahrtangelegenheiten (DARA) under grant 50 OR and the Netherlands Organization for Scientific Research (NWO) under grant PGS

8 8 A. van Teeseling et al.: The X-ray source in non-magnetic cataclysmic variables References Baykal, A., Esendemir, A., Kiziloglu, Ü., Alpar, M. A., Ögelman, H., Ercan, N., and Ikis, G., 1995, A&A 299, 421 Belloni, T., Verbunt, F., Beuermann, K., Bunk, W., Izzo, C., Kley, W., Pietsch, W., Ritter, H., Thomas, H. C., and Voges, W., 1991, A&A 246, L44 Beuermann, K. and Thomas, H.-C. 1993, Adv. Sp. Res. 13, (12)115 Kaastra, J. S. and Mewe, R. 1993, Legacy 3, 16 Mason, K. O., Lampton, M., Charles, P., and Bowyer, S., 1978, ApJ 226, L129 Mauche, C. W., 1995, in Cataclysmic Variables and Related Objects, IAU Coll. 158, eds. A. Evans & J. H. Wood, Kluwer Dordrecht Mauche, C. W., Raymond, J. C., and Mattei, J. A., 1995, ApJ 446, 842 Mukai, K., Schlegel, E. M., Swank, J. H., Naylor, T., and Wood, J. H., 1995, in Cataclysmic Variables and Related Objects, IAU Coll. 158, eds. A. Evans & J. H. Wood, Kluwer Academic Press, Dordrecht Naylor, T., Bath, G. T., Charles, P. A., Hassall, B. J. M., Sonneborn, G., Van der Woerd, H., and Van Paradijs, J., 1988, MNRAS 231, 237 Patterson, J., 1984, ApJS 54, 443 Patterson, J. and Raymond, J. C., 1985, ApJ 292, 535 Patterson, J., Sterner, E., Halpern, J. P., and Raymond, J. C., 1992, ApJ 384, 234 Patterson, J., 1994, PASP 106, 209 Ponman, T. J., Belloni, T., Duck, S. R., Verbunt, F., Watson, M. G., Wheatley, P. J., and Pfeffermann, E., 1995, MNRAS 276, 495 Predehl, P. and Schmitt, J. H. M. M., 1995, A&A 293, 889 Pringle, J. E. and Savonije, G. J., 1979, MNRAS 187, 777 Provencal, J. L., Winget, D. E., Nather, R. E., Robinson, E. L., Solheim, J.-E., et al., 1995, ApJ 445, 927 Rappaport, S., Cash, W., Doxsey, R., McClintock, J., and Moore, G., 1974, ApJ 187, L5 Richman, H. R. and Patterson, J., 1994, BAAS 26, 1345 Ritter, H. and Kolb, U., 1993, MPA preprint 757 Robinson, C. R. and Córdova, F. A., 1992, BAAS 24, 772 Rosen, S. R., Clayton, K. L., Osborne, J. P., and McGale, P. A., 1994, MNRAS 269, 913 Rosen, S. R., Watson T. K., Robinson E. L., Prinja R. K., Misselt K., Shafter A. W., 1995, A&A 300, 392 Schlegel, E. M. and Singh, J., 1995, MNRAS 276, 1365 Shafter, A. W., Szkody, P., and Thorstensen, J. R., 1986, ApJ 308, 765 Shafter, A. W., Robinson, E. L., Crampton, D., Warner, B., and Prestage, R. M., 1990, ApJ 354, 708 Silber, A., Vrtilek, S. D., and Raymond, J. C., 1994, ApJ 425, 829 Szkody, P., 1986, ApJ 301, L29 Tylenda, R., 1981, Acta. Astr. 31, 267 Ulla, A., 1995, A&A 301, 469 Van der Woerd, H. J., 1987, PhD thesis, University of Amsterdam Van der Woerd, H. and Heise, J., 1987, MNRAS 225, 141 Van Teeseling, A., Verbunt, F., and Heise, J., 1993, A&A 270, 159 Van Teeseling, A. and Verbunt, F., 1994, A&A, 292, 519 Van Teeseling, A., Drake, J. J., Drew, J. E., Hoare, M. G., and Verbunt, F., 1995, A&A 300, 808 Verbunt, F., 1987, A&AS 71, 339 Vila, C. S., 1971, ApJ 168, 217 Vrtilek, S. D., Silber, A., Raymond, J. C., and Patterson, J., 1994, ApJ 425, 787 Warner, B., 1987, MNRAS 227, 23 Wheatley, P. J., Verbunt, F., Belloni, T., Watson, M. G., Naylor, T., Ishida, M., Duck, S. R., and Pfeffermann, E., 1995, A&A 307, 137 Wheatley, P. J., Van Teeseling, A., Watson, M. G., Verbunt, F., and Pfeffermann., E., 1996, MNRAS, submitted Wood, J. H., Naylor, T., Hassall, B. J. M., and Ramseyer, T. F., 1995a, MNRAS 273, 772 (a) Wood, J. H., Naylor, T., and Marsh, T. R., 1995b, MNRAS 274, 31 Zimmermann, H. U., Becker, W., Belloni, T., Döbereiner, S., Izzo, C., Kahabka, P., and Schwentker, O., 1994, EXSAS User s Guide, MPE Report 257 This article was processed by the author using Springer-Verlag LaT E X A&A style file L-AA version 3.